Benoit Steiner
0360c36170
Merged in codeplaysoftware/eigen-upstream-pure/separating_internal_memory_allocation (pull request PR-446)
...
Distinguishing between internal memory allocation/deallocation from explicit user memory allocation/deallocation.
2018-08-01 16:13:15 +00:00
Mehdi Goli
b512a9536f
Enabling per device specialisation of packetsize.
2018-08-01 13:39:13 +01:00
Mehdi Goli
d7a8414848
Distinguishing between internal memory allocation/deallocation from explicit user memory allocation/deallocation.
2018-08-01 11:56:30 +01:00
Eugene Zhulenev
6913221c43
Add tiled evaluation support to TensorExecutor
2018-07-25 13:51:10 -07:00
Deven Desai
876f392c39
Updates corresponding to the latest round of PR feedback
...
The major changes are
1. Moving CUDA/PacketMath.h to GPU/PacketMath.h
2. Moving CUDA/MathFunctions.h to GPU/MathFunction.h
3. Moving CUDA/CudaSpecialFunctions.h to GPU/GpuSpecialFunctions.h
The above three changes effectively enable the Eigen "Packet" layer for the HIP platform
4. Merging the "hip_basic" and "cuda_basic" unit tests into one ("gpu_basic")
5. Updating the "EIGEN_DEVICE_FUNC" marking in some places
The change has been tested on the HIP and CUDA platforms.
2018-07-11 10:39:54 -04:00
Deven Desai
b6cc0961b1
updates based on PR feedback
...
There are two major changes (and a few minor ones which are not listed here...see PR discussion for details)
1. Eigen::half implementations for HIP and CUDA have been merged.
This means that
- `CUDA/Half.h` and `HIP/hcc/Half.h` got merged to a new file `GPU/Half.h`
- `CUDA/PacketMathHalf.h` and `HIP/hcc/PacketMathHalf.h` got merged to a new file `GPU/PacketMathHalf.h`
- `CUDA/TypeCasting.h` and `HIP/hcc/TypeCasting.h` got merged to a new file `GPU/TypeCasting.h`
After this change the `HIP/hcc` directory only contains one file `math_constants.h`. That will go away too once that file becomes a part of the HIP install.
2. new macros EIGEN_GPUCC, EIGEN_GPU_COMPILE_PHASE and EIGEN_HAS_GPU_FP16 have been added and the code has been updated to use them where appropriate.
- `EIGEN_GPUCC` is the same as `(EIGEN_CUDACC || EIGEN_HIPCC)`
- `EIGEN_GPU_DEVICE_COMPILE` is the same as `(EIGEN_CUDA_ARCH || EIGEN_HIP_DEVICE_COMPILE)`
- `EIGEN_HAS_GPU_FP16` is the same as `(EIGEN_HAS_CUDA_FP16 or EIGEN_HAS_HIP_FP16)`
2018-06-14 10:21:54 -04:00
Deven Desai
8fbd47052b
Adding support for using Eigen in HIP kernels.
...
This commit enables the use of Eigen on HIP kernels / AMD GPUs. Support has been added along the same lines as what already exists for using Eigen in CUDA kernels / NVidia GPUs.
Application code needs to explicitly define EIGEN_USE_HIP when using Eigen in HIP kernels. This is because some of the CUDA headers get picked up by default during Eigen compile (irrespective of whether or not the underlying compiler is CUDACC/NVCC, for e.g. Eigen/src/Core/arch/CUDA/Half.h). In order to maintain this behavior, the EIGEN_USE_HIP macro is used to switch to using the HIP version of those header files (see Eigen/Core and unsupported/Eigen/CXX11/Tensor)
Use the "-DEIGEN_TEST_HIP" cmake option to enable the HIP specific unit tests.
2018-06-06 10:12:58 -04:00
Gael Guennebaud
bbd97b4095
Add a EIGEN_NO_CUDA option, and introduce EIGEN_CUDACC and EIGEN_CUDA_ARCH aliases
2017-07-17 01:02:51 +02:00
Benoit Steiner
53725c10b8
Merged in mehdi_goli/opencl/DataDependancy (pull request PR-10)
...
DataDependancy
* Wrapping data type to the pointer class for sycl in non-terminal nodes; not having that breaks Tensorflow Conv2d code.
* Applying Ronnan's Comments.
* Applying benoit's comments
2017-06-28 17:55:23 +00:00
Benoit Steiner
66c63826bd
Guard the sycl specific code with EIGEN_USE_SYCL
2017-04-04 09:59:09 -07:00
Mehdi Goli
bd64ee8555
Fixing TensorArgMaxSycl.h; Removing warning related to the hardcoded type of dims to be int in Argmax.
2017-03-28 16:50:34 +01:00
Benoit Steiner
f0f3591118
Made the reduction code compile with cuda-clang
2017-03-14 14:16:53 -07:00
Mehdi Goli
35bae513a0
Converting all parallel for lambda to functor in order to prevent kernel duplication name error; adding tensorConcatinationOp backend for sycl.
2016-12-16 19:46:45 +00:00
Mehdi Goli
7318daf887
Fixing LLVM error on TensorMorphingSycl.h on GPU; fixing int64_t crash for tensor_broadcast_sycl on GPU; adding get_sycl_supported_devices() on syclDevice.h.
2016-11-25 16:19:07 +00:00
Benoit Steiner
dcc14bee64
Fixed the formatting of the code
2016-11-08 14:24:46 -08:00
Mehdi Goli
d57430dd73
Converting all sycl buffers to uninitialised device only buffers; adding memcpyHostToDevice and memcpyDeviceToHost on syclDevice; modifying all examples to obey the new rules; moving sycl queue creating to the device based on Benoit suggestion; removing the sycl specefic condition for returning m_result in TensorReduction.h according to Benoit suggestion.
2016-11-08 17:08:02 +00:00
Mehdi Goli
0ebe3808ca
Removed the sycl include from Eigen/Core and moved it to Unsupported/Eigen/CXX11/Tensor; added TensorReduction for sycl (full reduction and partial reduction); added TensorReduction test case for sycl (full reduction and partial reduction); fixed the tile size on TensorSyclRun.h based on the device max work group size;
2016-11-04 18:18:19 +00:00
Mehdi Goli
e36cb91c99
Fixing the code indentation in the TensorReduction.h file.
2016-10-14 18:03:00 +01:00
Mehdi Goli
524fa4c46f
Reducing the code by generalising sycl backend functions/structs.
2016-10-14 12:09:55 +01:00
Benoit Steiner
028e299577
Fixed a bug impacting some outer reductions on GPU
2016-09-12 18:36:52 -07:00
Benoit Steiner
64e68cbe87
Don't attempt to optimize partial reductions when the optimized implementation doesn't buy anything.
2016-08-08 19:29:59 -07:00
Benoit Steiner
c6b0de2c21
Improved partial reductions in more cases
2016-07-22 17:18:20 -07:00
Gael Guennebaud
544935101a
Fix warnings
2016-07-08 11:38:52 +02:00
Gael Guennebaud
179ebb88f9
Fix warning
2016-07-07 09:16:40 +02:00
Benoit Steiner
c21eaedce6
Use array_prod to compute the number of elements contained in the input tensor expression
2016-06-04 07:47:04 -07:00
Benoit Steiner
c2a102345f
Improved the performance of full reductions.
...
AFTER:
BM_fullReduction/10 4541 4543 154017 21.0M items/s
BM_fullReduction/64 5191 5193 100000 752.5M items/s
BM_fullReduction/512 9588 9588 71361 25.5G items/s
BM_fullReduction/4k 244314 244281 2863 64.0G items/s
BM_fullReduction/5k 359382 359363 1946 64.8G items/s
BEFORE:
BM_fullReduction/10 9085 9087 74395 10.5M items/s
BM_fullReduction/64 9478 9478 72014 412.1M items/s
BM_fullReduction/512 14643 14646 46902 16.7G items/s
BM_fullReduction/4k 260338 260384 2678 60.0G items/s
BM_fullReduction/5k 385076 385178 1818 60.5G items/s
2016-06-03 17:27:08 -07:00
Benoit Steiner
36369ab63c
Resolved merge conflicts
2016-05-26 13:39:39 -07:00
Benoit Steiner
28fcb5ca2a
Merged latest reduction improvements
2016-05-26 12:19:33 -07:00
Benoit Steiner
c1c7f06c35
Improved the performance of inner reductions.
2016-05-26 11:53:59 -07:00
Benoit Steiner
a09cbf9905
Merged in rmlarsen/eigen (pull request PR-188)
...
Minor cleanups: 1. Get rid of a few unused variables. 2. Get rid of last uses of EIGEN_USE_COST_MODEL.
2016-05-23 12:55:12 -07:00
Gael Guennebaud
ccaace03c9
Make EIGEN_HAS_CONSTEXPR user configurable
2016-05-20 15:10:08 +02:00
Gael Guennebaud
c3410804cd
Make EIGEN_HAS_VARIADIC_TEMPLATES user configurable
2016-05-20 15:05:38 +02:00
Rasmus Munk Larsen
7df811cfe5
Minor cleanups: 1. Get rid of unused variables. 2. Get rid of last uses of EIGEN_USE_COST_MODEL.
2016-05-18 15:09:48 -07:00
Benoit Steiner
8d06c02ffd
Allow vectorized padding on GPU. This helps speed things up a little.
...
Before:
BM_padding/10 5000000 460 217.03 MFlops/s
BM_padding/80 5000000 460 13899.40 MFlops/s
BM_padding/640 5000000 461 888421.17 MFlops/s
BM_padding/4K 5000000 460 54316322.55 MFlops/s
After:
BM_padding/10 5000000 454 220.20 MFlops/s
BM_padding/80 5000000 455 14039.86 MFlops/s
BM_padding/640 5000000 452 904968.83 MFlops/s
BM_padding/4K 5000000 411 60750049.21 MFlops/s
2016-05-17 09:13:27 -07:00
Benoit Steiner
09653e1f82
Improved the portability of the tensor code
2016-05-11 23:29:09 -07:00
Benoit Steiner
4ede059de1
Properly gate the use of half2.
2016-05-10 17:04:01 -07:00
Benoit Steiner
4670d7d5ce
Improved the performance of full reductions on GPU:
...
Before:
BM_fullReduction/10 200000 11751 8.51 MFlops/s
BM_fullReduction/80 5000 523385 12.23 MFlops/s
BM_fullReduction/640 50 36179326 11.32 MFlops/s
BM_fullReduction/4K 1 2173517195 11.50 MFlops/s
After:
BM_fullReduction/10 500000 5987 16.70 MFlops/s
BM_fullReduction/80 200000 10636 601.73 MFlops/s
BM_fullReduction/640 50000 58428 7010.31 MFlops/s
BM_fullReduction/4K 1000 2006106 12461.95 MFlops/s
2016-05-09 17:09:54 -07:00
Rasmus Munk Larsen
07ac4f7e02
Eigen Tensor cost model part 2: Thread scheduling for standard evaluators and reductions. The cost model is turned off by default.
2016-04-14 18:28:23 -07:00
Rasmus Munk Larsen
235e83aba6
Eigen cost model part 1. This implements a basic recursive framework to estimate the cost of evaluating tensor expressions.
2016-04-14 13:57:35 -07:00
Benoit Steiner
1bc81f7889
Fixed compilation warnings on arm
2016-03-28 09:21:04 -07:00
Benoit Steiner
41434a8a85
Avoid unnecessary conversions
2016-03-23 16:52:38 -07:00
Benoit Steiner
92693b50eb
Fixed compilation warning
2016-03-23 16:40:36 -07:00
Benoit Steiner
002cf0d1c9
Use a single Barrier instead of a collection of Notifications to reduce the thread synchronization overhead
2016-03-22 15:24:23 -07:00
Benoit Steiner
3149b5b148
Avoid implicit cast
2016-03-09 17:35:17 -08:00
Benoit Steiner
f05fb449b8
Avoid unnecessary conversion from 32bit int to 64bit unsigned int
2016-03-09 15:27:45 -08:00
Benoit Steiner
46177c8d64
Replace std::vector with our own implementation, as using the stl when compiling with nvcc and avx enabled leads to many issues.
2016-03-08 16:37:27 -08:00
Benoit Steiner
6d6413f768
Simplified the full reduction code
2016-03-08 16:02:00 -08:00
Benoit Steiner
e09eb835db
Decoupled the packet type definition from the definition of the tensor ops. All the vectorization is now defined in the tensor evaluators. This will make it possible to relialably support devices with different packet types in the same compilation unit.
2016-03-08 12:07:33 -08:00
Benoit Steiner
b2075cb7a2
Made the signature of the inner and outer reducers consistent
2016-02-29 10:53:38 -08:00
Benoit Steiner
3284842045
Optimized the performance of narrow reductions on CUDA devices
2016-02-29 10:48:16 -08:00
